Study of High Efficiency Video Coding Technologies for Cloud Gaming
DescriptionWith the advent of 5G era, cloud games are gaining more and more attention. Since the real-time rendering of game scenes requires high computational resources, cloud games can realize real-time rendering of gaming scenes on the server side and subsequently transmit the rendered video stream by video compression to lightweight mobile ends. In this way, users can receive high-quality gaming frames without equating high-performance processors or graphics cards. The delay of the cloud game video at the server-side during compression or end-to-end delay during transmission may introduce stalls, jumps, and visual quality degradation, which significantly influence the user quality of experience. Improving the quality of experience by promoting the consistency, fluency, and quality of game scenes are primary concerns in cloud gaming research and deployment. In the cloud game system, since the game video rendered by the server needs to be transmitted to the client in an efficient, extremely low-latency, and high-quality manner, how to efficiently encode the cloud game video is the core of the cloud game system, which is also the focus of this project. This project plans to improve the service quality of cloud games from three aspects. Firstly, we plan to establish a rate-distortion-complexity model to achieve low-complexity and low-latency cloud-game video coding. Secondly, the reinforcement learning based efficient rate control techniques are further developed. Finally, a high-quality game video coding and reconstruction framework will be investigated, thereby improving the ultimate user experience of cloud game applications.
|Effective start/end date||1/04/21 → …|